深度学习在医学图像上应用

细胞病理学识别和疾病组织检测是目标人工智能在影像医学和病理方向的重要应用。该技术主要是前期预处理技术复杂，原因是因为医学病理特征成因复杂，图像随机误差很大（噪音），断层之间的重叠。如传统图像的颜色没有识别要求，为了训练和计算方便采用的降维处理是将其灰化处理，然后使用分割算法将形态分离，该过程包含先将腐蚀再膨胀然后过滤（一般情况中值滤波能去掉白色噪音和边缘干扰），最后使用分类和回归来达到识别。深度学习方法一般如下：
Object Detection单目标检测和多目标检测

医学图像检测（CT/MRI）
1993年，CNN应用于肺结节检测；1995年CNN用于检测乳腺摄影中的微钙化。

医学图像分割：
如基于统计学的方法、
基于模糊理论的方法、
基于神经网络的方法、
基于小波分析的方法、
基于动态轮廓的方法、
组合优化模型等方法。

Unet+++
论文地址：https://arxiv.org/pdf/1807.10165.pdf
代码地址：https://github.com/MrGiovanni/UNetPlusPlus





脑肿瘤图像分割

CT肺部图像分割

Unet遥感图像分割

GAN图像分割

RCNN
supp: http://people.eecs.berkeley.edu/~rbg/papers/r-cnn-cvpr-supp.pdf
slides: http://www.image-net.org/challenges/LSVRC/2013/slides/r-cnn-ilsvrc2013-workshop.pdf
slides: http://www.cs.berkeley.edu/~rbg/slides/rcnn-cvpr14-slides.pdf
github: https://github.com/rbgirshick/rcnn

Fast R-CNN
arxiv: http://arxiv.org/abs/1504.08083
slides: http://tutorial.caffe.berkeleyvision.org/caffe-cvpr15-detection.pdf
github: https://github.com/rbgirshick/fast-rcnn
github: https://github.com/mahyarnajibi/fast-rcnn-torch
github: https://github.com/apple2373/chainer-simple-fast-rnn

Faster R-CNN
intro: NIPS 2015
arxiv: http://arxiv.org/abs/1506.01497
gitxiv: http://www.gitxiv.com/posts/8pfpcvefDYn2gSgXk/faster-r-cnn-towards-real-time-object-detection-with-region
slides: http://web.cs.hacettepe.edu.tr/~aykut/classes/spring2016/bil722/slides/w05-FasterR-CNN.pdf
github(official, Matlab): https://github.com/ShaoqingRen/faster_rcnn
github(Caffe): https://github.com/rbgirshick/py-faster-rcnn
github(MXNet): https://github.com/msracver/Deformable-ConvNets/tree/master/faster_rcnn
github: https://github.com/mitmul/chainer-faster-rcnn
github(PyTorch):: https://github.com/andreaskoepf/faster-rcnn.torch
github(PyTorch):: https://github.com/ruotianluo/Faster-RCNN-Densecap-torch
github(TensorFlow): https://github.com/smallcorgi/Faster-RCNN_TF

Light Head R-CNN
A Fast R-CNN
intro: CVPR 2017
arxiv: https://arxiv.org/abs/1704.03414
paper: http://abhinavsh.info/papers/pdfs/adversarial_object_detection.pdf
github(Caffe): https://github.com/xiaolonw/adversarial-frcnn

R-CNN minus R
intro: BMVC 2015
arxiv: http://arxiv.org/abs/1506.06981
Faster R-CNN in MXNet with distributed implementation and data parallelization
github: https://github.com/dmlc/mxnet/tree/master/example/rcnn

Contextual Priming and Feedback for Faster R-CNN
intro: ECCV 2016. Carnegie Mellon University
paper: http://abhinavsh.info/context_priming_feedback.pdf
poster: http://www.eccv2016.org/files/posters/P-1A-20.pdf

Interpretable R-CNN
intro: North Carolina State University & Alibaba
keywords: AND-OR Graph (AOG)
arxiv: https://arxiv.org/abs/1711.05226

Cascade R-CNN
MultiBox
SPP-Net
intro: ECCV 2014 / TPAMI 2015
arxiv: http://arxiv.org/abs/1406.4729
github: https://github.com/ShaoqingRen/SPP_net
notes: http://zhangliliang.com/2014/09/13/paper-note-sppnet/

MR—CNN
DeepID-Net: Deformable Deep Convolutional Neural Networks for Object Detection
intro: PAMI 2016
intro: an extension of R-CNN. box pre-training, cascade on region proposals, deformation layers and context representations
project page: http://www.ee.cuhk.edu.hk/%CB%9Cwlouyang/projects/imagenetDeepId/index.html
arxiv: http://arxiv.org/abs/1412.5661

Object Detectors Emerge in Deep Scene CNNs
intro: ICLR 2015
arxiv: http://arxiv.org/abs/1412.6856
paper: https://www.robots.ox.ac.uk/~vgg/rg/papers/zhou_iclr15.pdf
paper: https://people.csail.mit.edu/khosla/papers/iclr2015_zhou.pdf
slides: http://places.csail.mit.edu/slide_iclr2015.pdf

SegDeepM: Exploiting Segmentation and Context in Deep Neural Networks for Object Detection
intro: CVPR 2015
project(code+data): https://www.cs.toronto.edu/~yukun/segdeepm.html
arxiv: https://arxiv.org/abs/1502.04275
github: https://github.com/YknZhu/segDeepM

Object Detection Networks on Convolutional Feature Maps
intro: TPAMI 2015
keywords: NoC
arxiv: http://arxiv.org/abs/1504.06066
Improving Object Detection with Deep Convolutional Networks via

Bayesian Optimization and Structured Prediction
arxiv: http://arxiv.org/abs/1504.03293
slides: http://www.ytzhang.net/files/publications/2015-cvpr-det-slides.pdf
github: https://github.com/YutingZhang/fgs-obj

DeepBox: Learning Objectness with Convolutional Networks
keywords: DeepBox
arxiv: http://arxiv.org/abs/1505.02146
github: https://github.com/weichengkuo/DeepBox

LightNet: Bringing pjreddie’s DarkNet out of the shadows
https://github.com//explosion/lightnet

YOLO v2 Bounding Box Tool
intro: Bounding box labeler tool to generate the training data in the format YOLO v2 requires.
github: https://github.com/Cartucho/yolo-boundingbox-labeler-GUI

Loss Rank Mining: A General Hard Example Mining Method for Real-time Detectors
intro: LRM is the first hard example mining strategy which could fit YOLOv2 perfectly and make it better applied in series of real scenarios where both real-time rates and accurate detection are strongly demanded.
arxiv: https://arxiv.org/abs/1804.04606
YOLO3
YOLOv3: An Incremental Improvement
arxiv:https://arxiv.org/abs/1804.02767
paper:https://pjreddie.com/media/files/papers/YOLOv3.pdf
code: https://pjreddie.com/darknet/yolo/
github(Official):https://github.com/pjreddie/darknet
github:https://github.com/experiencor/keras-yolo3
github:https://github.com/qqwweee/keras-yolo3
github:https://github.com/marvis/pytorch-yolo3
github:https://github.com/ayooshkathuria/pytorch-yolo-v3
github:https://github.com/ayooshkathuria/YOLO_v3_tutorial_from_scratch

DenseBox
SSD
intro: ECCV 2016 Oral
arxiv: http://arxiv.org/abs/1512.02325
paper: http://www.cs.unc.edu/~wliu/papers/ssd.pdf
slides: http://www.cs.unc.edu/%7Ewliu/papers/ssd_eccv2016_slide.pdf
github(Official): https://github.com/weiliu89/caffe/tree/ssd
video: http://weibo.com/p/2304447a2326da963254c963c97fb05dd3a973
github: https://github.com/zhreshold/mxnet-ssd
github: https://github.com/zhreshold/mxnet-ssd.cpp
github: https://github.com/rykov8/ssd_keras
github: https://github.com/balancap/SSD-Tensorflow
github: https://github.com/amdegroot/ssd.pytorch
github(Caffe): https://github.com/chuanqi305/MobileNet-SSD
What’s the diffience in performance between this new code you pushed and the previous code? #327
https://github.com/weiliu89/caffe/issues/327

DSSD
DSSD : Deconvolutional Single Shot Detector
intro: UNC Chapel Hill & Amazon Inc
arxiv: https://arxiv.org/abs/1701.06659
github: https://github.com/chengyangfu/caffe/tree/dssd
github: https://github.com/MTCloudVision/mxnet-dssd
demo: http://120.52.72.53/www.cs.unc.edu/c3pr90ntc0td/~cyfu/dssd_lalaland.mp4

Enhancement of SSD by concatenating feature maps for object detection
intro: rainbow SSD (R-SSD)
arxiv: https://arxiv.org/abs/1705.09587

Context-aware Single-Shot Detector
keywords: CSSD, DiCSSD, DeCSSD, effective receptive fields (ERFs), theoretical receptive fields (TRFs)
arxiv: https://arxiv.org/abs/1707.08682
Feature-Fused SSD: Fast Detection for Small Objects
https://arxiv.org/abs/1709.05054

FSSD
FSSD: Feature Fusion Single Shot Multibox Detector
https://arxiv.org/abs/1712.00960

Weaving Multi-scale Context for Single Shot Detector
intro: WeaveNet
keywords: fuse multi-scale information
arxiv: https://arxiv.org/abs/1712.03149
ESSD
Extend the shallow part of Single Shot MultiBox Detector via Convolutional Neural Network
https://arxiv.org/abs/1801.05918

Tiny SSD: A Tiny Single-shot Detection Deep Convolutional Neural Network for Real-time Embedded Object Detection
https://arxiv.org/abs/1802.06488

Pelee
Pelee: A Real-Time Object Detection System on Mobile Devices
https://github.com/Robert-JunWang/Pelee
intro: (ICLR 2018 workshop track)
arxiv: https://arxiv.org/abs/1804.06882
github: https://github.com/Robert-JunWang/Pelee

R-FCN
R-FCN: Object Detection via Region-based Fully Convolutional Networks
arxiv: http://arxiv.org/abs/1605.06409
github: https://github.com/daijifeng001/R-FCN
github(MXNet): https://github.com/msracver/Deformable-ConvNets/tree/master/rfcn
github: https://github.com/Orpine/py-R-FCN
github: https://github.com/PureDiors/pytorch_RFCN
github: https://github.com/bharatsingh430/py-R-FCN-multiGPU
github: https://github.com/xdever/RFCN-tensorflow

R-FCN-3000 at 30fps: Decoupling Detection and Classification
https://arxiv.org/abs/1712.01802

Recycle deep features for better object detection
arxiv: http://arxiv.org/abs/1607.05066

FPN
Feature Pyramid Networks for Object Detection
intro: Facebook AI Research
arxiv: https://arxiv.org/abs/1612.03144

Action-Driven Object Detection with Top-Down Visual Attentions
arxiv: https://arxiv.org/abs/1612.06704

Beyond Skip Connections: Top-Down Modulation for Object Detection
intro: CMU & UC Berkeley & Google Research
arxiv: https://arxiv.org/abs/1612.06851

Wide-Residual-Inception Networks for Real-time Object Detection
intro: Inha University
arxiv: https://arxiv.org/abs/1702.01243

Attentional Network for Visual Object Detection
intro: University of Maryland & Mitsubishi Electric Research Laboratories
arxiv: https://arxiv.org/abs/1702.01478

Learning Chained Deep Features and Classifiers for Cascade in Object Detection
keykwords: CC-Net
intro: chained cascade network (CC-Net). 81.1% mAP on PASCAL VOC 2007
arxiv: https://arxiv.org/abs/1702.07054

DeNet: Scalable Real-time Object Detection with Directed Sparse Sampling
intro: ICCV 2017 (poster)
arxiv: https://arxiv.org/abs/1703.10295

Discriminative Bimodal Networks for Visual Localization and Detection with Natural Language Queries
intro: CVPR 2017
arxiv: https://arxiv.org/abs/1704.03944

Spatial Memory for Context Reasoning in Object Detection
arxiv: https://arxiv.org/abs/1704.04224

Accurate Single Stage Detector Using Recurrent Rolling Convolution
intro: CVPR 2017. SenseTime
keywords: Recurrent Rolling Convolution (RRC)
arxiv: https://arxiv.org/abs/1704.05776
github: https://github.com/xiaohaoChen/rrc_detection

Deep Occlusion Reasoning for Multi-Camera Multi-Target Detection
https://arxiv.org/abs/1704.05775

LCDet: Low-Complexity Fully-Convolutional Neural Networks for Object Detection in Embedded Systems
intro: Embedded Vision Workshop in CVPR. UC San Diego & Qualcomm Inc
arxiv: https://arxiv.org/abs/1705.05922

Point Linking Network for Object Detection
intro: Point Linking Network (PLN)
arxiv: https://arxiv.org/abs/1706.03646

Perceptual Generative Adversarial Networks for Small Object Detection
https://arxiv.org/abs/1706.05274

Few-shot Object Detection
https://arxiv.org/abs/1706.08249

Yes-Net: An effective Detector Based on Global Information
https://arxiv.org/abs/1706.09180

SMC Faster R-CNN: Toward a scene-specialized multi-object detector
https://arxiv.org/abs/1706.10217

Towards lightweight convolutional neural networks for object detection
https://arxiv.org/abs/1707.01395

RON: Reverse Connection with Objectness Prior Networks for Object Detection
intro: CVPR 2017
arxiv: https://arxiv.org/abs/1707.01691
github: https://github.com/taokong/RON

Mimicking Very Efficient Network for Object Detection
intro: CVPR 2017. SenseTime & Beihang University
paper: http://openaccess.thecvf.com/content_cvpr_2017/papers/Li_Mimicking_Very_Efficient_CVPR_2017_paper.pdf

Residual Features and Unified Prediction Network for Single Stage Detection
https://arxiv.org/abs/1707.05031

Deformable Part-based Fully Convolutional Network for Object Detection
intro: BMVC 2017 (oral). Sorbonne Universités & CEDRIC
arxiv: https://arxiv.org/abs/1707.06175

Adaptive Feeding: Achieving Fast and Accurate Detections by Adaptively Combining Object Detectors
intro: ICCV 2017
arxiv: https://arxiv.org/abs/1707.06399

Recurrent Scale Approximation for Object Detection in CNN
intro: ICCV 2017
keywords: Recurrent Scale Approximation (RSA)
arxiv: https://arxiv.org/abs/1707.09531
github: https://github.com/sciencefans/RSA-for-object-detection
DSOD
DSOD: Learning Deeply Supervised Object Detectors from Scratch

img
intro: ICCV 2017. Fudan University & Tsinghua University & Intel Labs China
arxiv: https://arxiv.org/abs/1708.01241
github: https://github.com/szq0214/DSOD
github:https://github.com/Windaway/DSOD-Tensorflow
github:https://github.com/chenyuntc/dsod.pytorch

Learning Object Detectors from Scratch with Gated Recurrent Feature Pyramids
arxiv:https://arxiv.org/abs/1712.00886
github:https://github.com/szq0214/GRP-DSOD

RetinaNet
Focal Loss for Dense Object Detection
intro: ICCV 2017 Best student paper award. Facebook AI Research
keywords: RetinaNet
arxiv: https://arxiv.org/abs/1708.02002

CoupleNet: Coupling Global Structure with Local Parts for Object Detection
intro: ICCV 2017
arxiv: https://arxiv.org/abs/1708.02863

Incremental Learning of Object Detectors without Catastrophic Forgetting
intro: ICCV 2017. Inria
arxiv: https://arxiv.org/abs/1708.06977

Zoom Out-and-In Network with Map Attention Decision for Region Proposal and Object Detection
https://arxiv.org/abs/1709.04347

StairNet: Top-Down Semantic Aggregation for Accurate One Shot Detection
https://arxiv.org/abs/1709.05788

Dynamic Zoom-in Network for Fast Object Detection in Large Images
https://arxiv.org/abs/1711.05187

Zero-Annotation Object Detection with Web Knowledge Transfer
intro: NTU, Singapore & Amazon
keywords: multi-instance multi-label domain adaption learning framework
arxiv: https://arxiv.org/abs/1711.05954

MegDet
MegDet: A Large Mini-Batch Object Detector
intro: Peking University & Tsinghua University & Megvii Inc
arxiv: https://arxiv.org/abs/1711.07240

Single-Shot Refinement Neural Network for Object Detection
arxiv: https://arxiv.org/abs/1711.06897
github: https://github.com/sfzhang15/RefineDet

Receptive Field Block Net for Accurate and Fast Object Detection
intro: RFBNet
arxiv: https://arxiv.org/abs/1711.07767
github: https://github.com//ruinmessi/RFBNet

An Analysis of Scale Invariance in Object Detection – SNIP
arxiv: https://arxiv.org/abs/1711.08189
github: https://github.com/bharatsingh430/snip

Feature Selective Networks for Object Detection
https://arxiv.org/abs/1711.08879

Learning a Rotation Invariant Detector with Rotatable Bounding Box
arxiv: https://arxiv.org/abs/1711.09405
github: https://github.com/liulei01/DRBox

Scalable Object Detection for Stylized Objects
intro: Microsoft AI & Research Munich
arxiv: https://arxiv.org/abs/1711.09822

Learning Object Detectors from Scratch with Gated Recurrent Feature Pyramids
arxiv: https://arxiv.org/abs/1712.00886
github: https://github.com/szq0214/GRP-DSOD

Deep Regionlets for Object Detection
keywords: region selection network, gating network
arxiv: https://arxiv.org/abs/1712.02408

Training and Testing Object Detectors with Virtual Images
intro: IEEE/CAA Journal of Automatica Sinica
arxiv: https://arxiv.org/abs/1712.08470

Large-Scale Object Discovery and Detector Adaptation from Unlabeled Video
keywords: object mining, object tracking, unsupervised object discovery by appearance-based clustering, self-supervised detector adaptation
arxiv: https://arxiv.org/abs/1712.08832

Spot the Difference by Object Detection
intro: Tsinghua University & JD Group
arxiv: https://arxiv.org/abs/1801.01051

Localization-Aware Active Learning for Object Detection
arxiv: https://arxiv.org/abs/1801.05124

Object Detection with Mask-based Feature Encoding
https://arxiv.org/abs/1802.03934

LSTD: A Low-Shot Transfer Detector for Object Detection
intro: AAAI 2018
arxiv: https://arxiv.org/abs/1803.01529

Domain Adaptive Faster R-CNN for Object Detection in the Wild
intro: CVPR 2018. ETH Zurich & ESAT/PSI
arxiv: https://arxiv.org/abs/1803.03243

Pseudo Mask Augmented Object Detection
https://arxiv.org/abs/1803.05858

Revisiting RCNN: On Awakening the Classification Power of Faster RCNN
https://arxiv.org/abs/1803.06799

Zero-Shot Detection
intro: Australian National University
keywords: YOLO
arxiv: https://arxiv.org/abs/1803.07113

Learning Region Features for Object Detection
intro: Peking University & MSRA
arxiv: https://arxiv.org/abs/1803.07066

Single-Shot Bidirectional Pyramid Networks for High-Quality Object Detection
intro: Singapore Management University & Zhejiang University
arxiv: https://arxiv.org/abs/1803.08208

Object Detection for Comics using Manga109 Annotations
intro: University of Tokyo & National Institute of Informatics, Japan
arxiv: https://arxiv.org/abs/1803.08670

Task-Driven Super Resolution: Object Detection in Low-resolution Images
https://arxiv.org/abs/1803.11316

Transferring Common-Sense Knowledge for Object Detection
https://arxiv.org/abs/1804.01077

Multi-scale Location-aware Kernel Representation for Object Detection
intro: CVPR 2018
arxiv: https://arxiv.org/abs/1804.00428
github: https://github.com/Hwang64/MLKP

Loss Rank Mining: A General Hard Example Mining Method for Real-time Detectors
intro: National University of Defense Technology
arxiv: https://arxiv.org/abs/1804.04606

Robust Physical Adversarial Attack on Faster R-CNN Object Detector
https://arxiv.org/abs/1804.05810

DetNet
DetNet: A Backbone network for Object Detection
intro: Tsinghua University & Face++
arxiv: https://arxiv.org/abs/1804.06215

Other
Relation Network for Object Detection
intro: CVPR 2018
arxiv: https://arxiv.org/abs/1711.11575

Quantization Mimic: Towards Very Tiny CNN for Object Detection
Tsinghua University1 & The Chinese University of Hong Kong2 &SenseTime3
arxiv: https://arxiv.org/abs/1805.02152
github：https://github.com/amusi/awesome-object-detection
Inside-OutsideNet（ION）
CRAFI
OHEM
R-FCN
MS-CNN
PVANET
GBD-Net
LocNet

理论基础上数字图像的形态学处理和识别
首先熟悉一下数字图像的定义和种类。数字图像：能在计算机上进行显示和操作的图像，图像是人的视觉系统对客观事物产生的视觉印象信号。按照格式常见的区分为：位图数字阵列BMP、JPG、GIF，矢量图像PNG（这种形式方便矩阵和向量（一种区别与数组的数据结构）。一般情况使用数字摄像机和数字照相机等电子设备采集的图像都认为是位图图像。由于时间问题今天先暂时更新这里：下面介绍一个入手项目仿照该项可以开发一套在线医疗检测系统

开发基础环境：tensorflow+keras来实现WEB端的多目标检测部署
YOLO_Online 将深度学习最火的目标检测做成在线医学服务
技术实现

web 用了 Django 来做界面，就是上传文件，保存文件这个功能。
YOLO 的实现用的是 keras-yolo3，直接导入yolo 官方的权重即可。
YOLO 和 web 的交互最后使用的是 socket。
问题一、
Django 中 Keras 初始化会有 bug，原计划是直接在 Django 里面用 keras，后来发现技术难度很大，最后用Django负责文件和socket把文件传给yolov3.
问题二、
最后 Django 是负责拿文件，然后用 socket 把文件名传给 yolo。
说好的在线服务，为什么没有上线呢？买了腾讯云 1 CPU 2 G 内存，部署的时候发现 keras 根本起不来，直接被 Killed 。
问题三、
解决，并没有解决，因为买不起更好地服务器了，只好本地运行然后截图了。通过本地测试完成了该项目。
django和yolo开发流程是没有把两个代码放一起，而是分开的。Django 只是上传文件，yolo 处理图片。
项目代码：链接:https://pan.baidu.com/s/1LNv2W8EBvuHKVE6npFsaGA 密码:nj2r

实现步骤：
1、从yolo官方网站下载yolo3的权重参数
https://link.zhihu.com/?target=https%3A//pjreddie.com/media/files/yolov3.weights
2、转换DarkNet模型位keras模型
将python convert.py yolo3.cfg yolov3.weights model_data/yolo.h5
3、运行yolo目标检测
python yolo.py
下载一张图片然后输入图的名称
该项目的源代码地址：https://github.com/qqwweee/keras-yolo3
该项目是Img训练的医学无法直接应用，需要通过迁移学习。
方法就是把模型和参数迁移到你自己的数据集上训练保存一个baseline，设计一个demo